Is it time to buy ASML?
Q2 Earnings Digest
ASML reported earnings earlier today. My expectations going into said earnings were not great in what concerns the market reaction because it wouldn’t be the first quarter in which the market makes an effort to fixate on any given number that’s pretty much irrelevant to the long-term thesis to justify a significant stock drop (this gave birth to the following meme):
Turns out it was quite a prescient meme, by the way:
Quarterly numbers were solid (beat on revenue, EPS, and quarterly orders) if we ignore the Q3 guide, which missed due to timing reasons. Management pointed out that they expect a “disproportionate” amount of sales in Q4 this year, but are targeting 15% growth for 2025, which exceeds the consensus of 13.8% growth (i.e., revenue is expected to be stronger than anticipated in Q4).
Orders were also significantly higher than consensus, so, without quarterly numbers to blame this time, the market began looking into the unclear 2026 commentary that management provided. Christophe Fouquet mentioned that they expect growth in 2026 but that they could not confirm it due to geopolitical uncertainty:
While we still prepare for growth in 2026, we cannot confirm it at this stage.
I’ll delve into this topic later on in the article. Without any other “warning signs” on the horizon, this 2026 commentary was what the market held onto to punish the stock. The constant search for a bearish thesis for the company is a good indicator of sentiment surrounding the stock. There’s some sort of overhang weighing on the stock currently due to several “issues”:
Sentiment is negative because ASML is not a short-term beneficiary of AI, unlike Nvidia or TSMC. It’s true that AI is probably benefiting ASML through higher IBM (Installed Base Management) sales as TSMC quickly ramps capacity, but the bulk of the sales are still generated by the equipment, which faces longer decision times by customers and longer production lead times
The “China risk” which has two angles to it. On the one hand, there are still concerns about China being able to develop an EUV competitor. Many people still believe that being able to generate EUV light is synonymous with having an EUV system working at scale. On the other hand, ASML is still deriving a significant portion of its sales from China, and some investors think that this business needs to come down to derisk its sales
Related to the above, we can find geopolitics. ASML is a key company in the semiconductor supply chain and is being used as a geopolitical weapon. While true, I don’t think this is a net negative, as there’s a side of the coin that calls for more domestic manufacturing that will benefit ASML over the medium to long term.
The industry is working on 3D structures to theoretically lower the dependence on lithography when manufacturing leading-edge chips.
Put all of these together and you likely get a negative drag on the stock so long as the numbers are not pristine. For someone looking for absolutely pristine numbers, the 2026 commentary and the fact that Q3 2025 guidance missed might be reason enough to remain bearish. Why someone would focus on quarterly guidance for a business where timing can significantly impact numbers after management beats consensus for the full year is beyond me, but I suppose that’s what makes a market.
Another interesting aspect is that, being more exposed to the capital expenditure (Capex) cycle of its customers, ASML is considered a cyclical company with limited visibility. While I am not here to deny the cyclicality of this business, I don’t think the financials and backlog portray ASML as a highly cyclical business. Should supply meet demand in the future and should lead times shorten significantly, I might change my opinion, but this is simply not the case today. Now, the fact that products retail for hundreds of millions of dollars does make timing an important consideration that some people might confuse with a lack of demand/cyclicality (I’ll discuss this later on).
Despite all the quarterly noise that’s generated every three months for a company with lead times >1 year, I always like to bring the quarterly table:
Pretty much all the numbers you see above are not meaningful to the long-term thesis. I have discussed this many times, but when you sell products worth over €200 million, even small changes in the recognition of those products can have a significant impact on any given quarter. If said quarter is Q4, then it can even impact the results of any given year.
A good example of this is the development of quarterly orders. With many investors viewing this metric as the “ultimate leading indicator of sales” (a claim that has never been substantiated), quarterly orders receive significant attention every quarter. The thing is that they seem to receive attention only when they are weak, not when they are strong. Quarterly bookings significantly beat expectations and grew sequentially:
The one thing I would highlight about the table above is that ASML still has €7 billion in cash, which the company will most likely use to continue repurchasing shares. Management repurchased €1.4 billion worth of shares during the quarter.
Now, setting aside the quarterly commentary (which I am never fond of, but less so for ASML due to its inherent characteristics), let’s discuss some other longer-term and arguably more relevant topics.
Adoption, productivity, and litho intensity
Although quarterly numbers can fluctuate due to timing, understanding the state of the company’s EUV business is crucial. Management shared some interesting comments here. First, they mentioned that low and high-NA EUV adoption remains on track, aided by significant productivity enhancements. Low-NA EUV is getting upgraded on the field to the NXE:3800 configuration, which comes with a 37% productivity improvement (productivity = wafers per hour). This has some relevant implications for system sales because (according to management) customer needs in terms of low-NA EUV have risen by 30%, but pretty much all of this has been covered with the upgrades. This ultimately means that sales of low-NA EUV units are expected to be flat this year. While this is a “blow” to system unit sales over the short term, it’s actually positive for ASML’s low-NA business due to higher ASPs and higher margins:
Well, I think we usually define the EUV capacity needed by the total need for wafer output. So I think this is the number we monitor. I think you have seen that in the Capital Market Day where we look at the total wafer start per month and how this evolve over time. So I think this define our market. And we try always to serve that market by shipping the most effective tool, the most productive tool, The reason for that is that you know, a tool with higher productivity will allow us to deliver the highest value to our customer with a lower cost on our side and therefore a better gross margin. So I think we are more sensitive to the total capacity need of our customer than to how many tool we're going to need to fulfill it. And I will say, the less number of tool we need to fulfill it, usually the better our margin and profitability will be. And for our customer, also the value is is is better because they have more space basically to run more capacity.
High-NA EUV is also experiencing productivity increases, which are even more relevant given the ongoing debate over whether multi-patterning low-NA EUV or single-patterning high-NA EUV is better (a key consideration in high-NA EUV adoption). A key variable to solve here is the technology cost. The NXE:5200 model brings a productivity improvement of 60% (up to 175 wph). The higher rate of improvement shouldn’t be a surprise considering high-NA’s lower level of maturity compared to low-NA EUV (which is already capable of 220 wph). The most interesting aspect of these productivity upgrades is how they are impacting litho intensity.
Management argued that increased productivity is leading to the conversion of multi-patterning DUV layers to single-exposed EUV layers, leading customers (including DRAM customers) to convert more layers into EUV. The rationale here is that, if single patterning is cheaper or at least comparable in cost to multi-patterning, you would most likely choose it as it reduces process complexity. This reduction in process complexity also allows customers to consider future nodes, which are expected to be more litho-intensive:
We explained that already a few times, this tool, thanks to the higher productivity, really allows customers to shift more multi-patterning layers to single exposure. So this basically allows customers to reduce complexity, reduce yield loss, improve cycle time. We have seen that happening quite a bit in the last few months with DRAM, where for the last nodes we really see a shift towards more EUV layers. This is, of course, in our case, a nice increase in litho intensity.
Everything starts with productivity, which is why comparisons between the cost/productivity of current high-NA systems and long-term litho intensity are a bit misleading. Sure, if high-NA’s cost never goes down, then it will be tough to displace low-NA EUV, but the reality is that productivity will rise and cost will go down, making high-NA EUV comparatively cheaper against low-NA multipatterning:
I think we are validating that better cost of technology allows us to translate more multi-patterning layers into single-exposed. I think we have made good progress on our litho intensity.
Of course, I have no idea where litho intensity will land over the long term, but I do know that taking current specifications into account to forecast it might not be the best idea.
The installed base management business
One thing I believe doesn’t get enough attention is ASML’s installed base management business. While ASML is well-known for its leading-edge lithography equipment, fewer people are aware of the longevity of this equipment and what it entails. Most of ASML’s systems are still operational, generating a substantial flow of upgrade and service revenue (which are not the same thing).
For example, ASML is currently upgrading its low-NA EUV NXE:3600 systems to the NXE:3800, which was a tailwind to both revenue and margin this quarter. Management expects all NXE:3600 to be converted to NXE:3800 this year. To this we must add that EUV tools are coming out of warranty, which should bode well for service revenue going forward.
Not much thought is given to this source of recurring revenue for ASML. The company’s equipment has very high longevity because chips have many layers that require different technologies, so we’ll most likely see IBM sales grow nicely as the installed base inevitably grows over the coming years.
What about the 2026 “miss”?
As discussed earlier, the likely most “worrying” part of the release was management’s commentary around the company preparing for 2026 to be a growth year, but their inability to confirm it due to geopolitical uncertainties. Some people took this as a warning sign, considering that ASML is typically touted for having high visibility due to the long lead times of its products (mainly EUV). I think these comments need to be contextualized.
First, we are still in Q2 2025, and there might be uncertainty related to the recognition of the backlog in 2026. Say ASML expects to ship two high-NA EUV systems in Q3/Q4 next year (totally making this up). If the environment is uncertain, then these two shipments might be pushed out to H1 2027, thereby depriving the company of around €800 million in revenue for 2026. While this is a made-up example, it should help us understand that having visibility into future demand doesn’t mean that management knows precisely when that demand will flow into the financials.
Also related to timing, we must be aware that ASML beat 2025 consensus growth expectations, meaning that a portion of the growth expected for 2026 might have been pulled forward. Analysts expected 7% growth for 2026, so the pull forward alone would not explain the entire differential, but it would explain a portion of it.
Secondly, we must not forget that the US is still in a “cold war” with China and ASML’s lithography equipment is a key weapon in this war. While the environment seems stable today, my good friend lokoyacap pointed out the following:
Gee I wonder why $asml would want to be conservative on 2026 right now despite things rn being solid
(buried in the WSJ article yesterday re the $nvda H20 approval, emphasis mine)
“The Commerce Department has prepared HARSHER restrictions on China, including limits on EQUIPMENT needed to make semiconductors, IN CASE trade relations sour again.”
In short, it makes little sense for the management team to be confident about 2026 numbers when they don’t even know if they will be available to ship said tools to China next year. Management mentioned that the H20 restrictions being lifted just yesterday are a good first step, but that customers are being cautious.
Medium to long term, I don’t see this as a problem because the 2030 guidance is based on global wafer demand, irrespective of where said wafers are manufactured. This not only applies to countries but also to customers. Intel and Samsung might be scaling back some of their expected Capex plans, given that TSMC has the lead and is taking all the volume. However, the demand will eventually have to be met, regardless of whether it’s TSMC, Samsung, or Intel that satisfies said demand. Of course, it would be much better for ASML to see all three customers thrive and compete against each other.
To this we must add that management is currently repurchasing shares and has €7 billion in the bank. Despite ASML not being the greatest repurchaser of its own stock, I do think this is an additional reason for not wanting to be extremely bullish amidst said uncertainty. So, all in all, I don’t think the 2026 is extremely worrying and it seems to be a case of management baking in conservativeness. Note that management put several “bear theses” to rest this quarter (albeit we still have to see how they pan out)...
Single patterning high-NA EUV is expected to be economical vs low-NA multi-patterning
Single patterning reduces process complexity and therefore increases litho intensity
DRAM customers are actively looking into EUV even as they transition to 3D structures